M5Stack M5StickV is a Tiny AI Camera for Maker Projects

Orange Pi Development Boards

I’ve just started to play with Maixduino board based on ESP32 WiSoC and Sipeed M1 module that enables AI tasks such as object detection thanks to built-in AI accelerators found in Kendryte K210 RISC-V processor and noticed references to M5Stack M5StickV in firmware file names.

Somehow I never wrote about M5Stack, but the company provides modular ESP32 IoT development boards that can be stacked with various modules to easily and quickly build prototypes. M5StickV is one of those modules and is similar to Maixduino kit with camera and display, minus WiFi + Bluetooth connectivity, except that everything nicely packed into a cute module.

M5StickV AI Camera

M5StickV hardware specifications:

  • SoC – Kendryte K210 dual-core 64-bit RISC-V processor @ 400MHz with dual independent double-precision FPU, 8MB on-chip SRAM, Neural Network Processor (KPU) @ 0.8Tops, Field-Programmable IO Array (FPIOA), and more
  • Storage – 16MB flash, microSD card slot
  • Display -1.14″ SPI display with 240×135 resolution ( ST7789 driver)
  • Camera – VGA (640×480) camera via OV7740 sensor
  • Audio – MAX98357 mono audio amplifier, MSM261S4030HOR microphone, speaker
  • Sensor – MPU6886 3-axis gyroscope, 3-axis accelerometer
  • USB – 1x USB-C port for power and programming
  • Misc – Front and side buttons (A / B), power button, RGBW LED
  • Expansion – 4-pin “CONNEXT” port
  • Power Supply
    • 200 mAh battery
    • AXP192 PMIC
  • Dimensions – Small… There’s also an M3 mounting thread

M5StickV DescriptionThe camera support OpenMV and MicroPython programming, and can be used for a variety of applications including face recognition/detection, object detection/classification, obtaining size and coordinates of the target in real-time, obtaining the type of detected target in real-time, shape recognition
video/audio recording, and game simulator. You’ll find instructions to get started in the Quick Start Guide.

M5StickV is sold “without WiFi” (so maybe a WiFi version is coming soon) on M5Stack website for $26.50, but you’ll also find it on Banggood for the same price.

Support CNX Software - Donate via PayPal or become a Patron on Patreon

9
Leave a Reply

avatar
3 Comment threads
6 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
4 Comment authors
Jean-Luc Aufranc (CNXSoft)Jon SmirltheguyukZombie Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
Zombie
Guest
Zombie

Jimmy’s hardware designer is testing a proto with wifi currently, so given how fast toarket Jimmy is, very soon

Member

Do we have any neural network benchmarking for this CPU on real-world models? For example I have a RK1808 on my desk running mobilenet-ssd object detection with 300x300px images. It can do 55FPS which is pretty good, better than my desktop CPU (no NN processor), of course my desktop GPU is way faster. You don’t really need to do object detection at 55FPS, 15FPS is fine. But what this lets you do is run a better model at 15FPS. Or you can run more than one model.

Zombie
Guest
Zombie

What kind of rk1808 product is that and where to get it?

Member

Pine64 and Radxa are both very close to releasing fairly cheap RK1808 dev boards. I am working with pre-release hardware from them.

Right now you can buy this USB compute stick (that Core i3 is a mistake, Core i3 is the minimum host requirement).
https://www.seeedstudio.com/RK1808-AI-Compute-Stick-Intel-Core-i3-Processor-p-4190.html

Or a module:
https://www.seeedstudio.com/BeiQi-RK1808-AIoT-96Boards-Compute-SoM-p-4074.html
https://www.seeedstudio.com/BeiQi-CarrierBoard-Kit-For-RK1808-RK3399Pro-p-4076.html

The upcoming Pine64 and Radxa solutions are less than half the price of these, but you can buy the these two right now. When Pine64 and Radxa ship they will be easy to order. My guess is maybe two months until they are for sale, but I don’t really know for sure.

Member

All of the RK1808 code is on github, no problem downloading the SDK.

theguyuk
Guest
theguyuk

Has the s905x3 not got a NPU built in?